"Many large language models in life science have now moved to a hyper-accelerated phase." – Eric Topol Recently, Stanford researchers introduced a "Virtual Lab"* of five domain-specific AI agents (see image), tasked them with designing nanobodies against SARS-CoV-2, and in record time, they delivered two potential candidates. It makes me wonder... What use cases in care delivery (not life science) could benefit from a Mixture of Experts (MoE) set-up? And how do we ensure clinicians are trained to leverage these experts effectively? Big questions, no answers from me yet (thinkering). *Source = The Virtual Lab: AI Agents Design New SARS-CoV-2 Nanobodies with Experimental Validation
Rik Renard’s Post
More Relevant Posts
-
Connor uses generative AI in a unique way: instead of applying it solely to propose new molecules, he leverages it to suggest experimental procedures. This approach optimizes the synthesis path by learning from existing chemical reaction data, enabling AI to recommend the most efficient steps for synthesizing a desired molecule. By proposing practical experimental routes, this method directly impacts the feasibility and speed of drug discovery.
Generative AI for Molecular Design & Synthesis
https://www.youtube.com/
-
Large Language Models could play a big role in reducing the burnout in healthcare industry where low risk tasks can be offloaded to these models. These presentations show the ongoing research in developing the new Foundational Models are evolving. Thanks to presenters also talk about the risks, accountability and share history of development of the models that are evaluated in the healthcare domain. Recordings from AIMI Symposium are available here : https://lnkd.in/gg6csf25 #stanfordaimi #healthcare Stanford University School of Medicine Stanford Center for Artificial Intelligence in Medicine and Imaging (AIMI)
-
What if AI can read Images from human thoughts. You can read research work from Meta on "Toward a real-time decoding of images from brain activity". https://lnkd.in/gHWVmgXY
Toward a real-time decoding of images from brain activity
ai.meta.com
-
enabling digital services for Student Loan related activities while maintaining the highest security standard, the most compliant personal data protection and customer-centric data-driven innovation.
I am pleased to announce the release of a new blog post titled "Discovering Sensorimotor Agency in Cellular Automata using Diversity Search" on arXiv:2402.10236v1. The post explores the potential for self-organization of "individuals" in cellular automata, utilizing recent advances in machine learning to automate the search for localized structures capable of reacting to external obstacles and maintaining integrity. The results indicate the discovery of agents with robust movement and generalization abilities, with implications for AI and synthetic bioengineering. Read the full post at https://bit.ly/3SLI5xh.
-
This week saw the release of Alphafold-3 and GPT-4o, advancing AI capabilities in protein prediction and multimodal interactions. Alphafold-3 can model folding patterns and chemical structures across biomolecules, while GPT-4o enables real-time interaction with speech, images, and video. These advancements have significant implications for drug development and AI accessibility.</div><div class="read-more"><a href="" class="more-link">Continue reading</a>https://lnkd.in/gqskMyVr
This AI newsletter is all you need #99
-
✨ A great example of how AI is transforming the Biotech industry: Profluent is leveraging a protein language model to develop CRISPR gene-editing proteins, broadening their range of editors with the help of AI. 🔬 And the best part? They're sharing their AI-generated OpenCRISPR-1 editor, which they've used to edit human DNA, as an open-source resource for anyone to experiment with! Read more in this Nature article: https://hubs.li/Q02yN9xb0
‘ChatGPT for CRISPR’ creates new gene-editing tools
nature.com
-
Kwaai AI Life Science research team, led by Diego Galeano are making excellent progress in improving the accuracy of their Drug Allergy Bot. Can AI accurately answer the question: "Is [symptom] an adverse reaction of [prescription drug]? LLM - 50% accurate (about as good as flipping a coin) LLM retrained on Drug Leaflet DB - 60% LLM + RAG - 80% LLM + GraphRAG - 98% Visit https://www.kwaai.ai/ to learn more Sign up to participate in the Kwaai, nonprofit AI Lab. #kwaai #ai #nonprofit #lab #lifescience
Boosting Language Model Precision with RAG, and GraphRAG 🚀 Diego Galeano reported on the progress of the Kwaai AI for Life Science workgroup. In this video, we dive into the innovative workflow of a retrieval augmented generation (RAG) system designed to improve the accuracy of language models. Discover how this technology processes data to enhance responses for questions about side effects in pharmacology. Kwaai is a nonprofit AI Lab. Learn more at https://www.kwaai.ai/ #Kwaai #ai #LanguageModel #TechInnovation #RACSystem #DataProcessing #AIAccuracy #MachineLearning #TechExploration #FutureOfAI #InformationRetrieval #AIEnhancements
-
with great (discriminatory/training) power comes great (oversight) responsibility...
Director, Centre for Population Genomics, Garvan Institute of Medical Research and Murdoch Children's Research Institute
AI exploding into biomedical applications at an incredibly rapid pace: “The amount of compute used to train biological sequence models has, over the past 6 years, increased by an average of 8-10 times per year. This rapid growth in training compute outpaces the rate of growth for computer vision and reinforcement learning systems, but is comparable to the growth rate for language models.” https://lnkd.in/g_i6mzxt
Biological Sequence Models in the Context of the AI Directives
epochai.org
-
Product Marketing Team Lead | Life Science | AI & Cheminformatics in Drug Discovery
🌟 Our generative-/predictive AI model, XMolGen, is here to accelerate your drug discovery decision-making by: 1️⃣ Drug-Like Novelty through De Novo Design: By leveraging generative AI and scaffold hopping, we design structurally unique compounds that optimize target engagement, strengthening your IP portfolio. 2️⃣ Comprehensive Multi-Parameter Optimization: Predictive models refine ADMET profiles to improve drug candidates' PK, minimize safety risk, and enhance clinical viability. 3️⃣ Synthetic Accessibility: We prioritize feasibility in our synthesis plans, enabling rapid wet lab validation and ensuring that designed compounds are ready for effective experimental evaluation and scalable production. Contact us for a free trial to experience our technology!! #DrugDiscovery #AI #GenerativeAI #XtalPi
Request Your Free Trial Now: https://hubs.li/Q02W9h1G0 In highly competitive intellectual property (IP) landscapes, existing methods like fragment-based design and high-throughput screening often fall short in generating compounds with both novelty and diversity. XtalPi’s XMolGen changes that, using AI and reinforcement learning to generate diverse scaffolds and explore new chemical spaces. With features like virtual docking, synthesis suggestions, and drug-like property predictions, XMolGen supports applications including de novo molecular generation, library enumeration, and virtual screening. Enhance novelty, strengthen your IP profile, and accelerate discovery with XMolGen. Free trials available upon request. #artificialintelligence #moleculargeneration #virtualscreening #drugdiscovery
-
-
Boosting Language Model Precision with RAG, and GraphRAG 🚀 Diego Galeano reported on the progress of the Kwaai AI for Life Science workgroup. In this video, we dive into the innovative workflow of a retrieval augmented generation (RAG) system designed to improve the accuracy of language models. Discover how this technology processes data to enhance responses for questions about side effects in pharmacology. Kwaai is a nonprofit AI Lab. Learn more at https://www.kwaai.ai/ #Kwaai #ai #LanguageModel #TechInnovation #RACSystem #DataProcessing #AIAccuracy #MachineLearning #TechExploration #FutureOfAI #InformationRetrieval #AIEnhancements
Physician | Founding Team @ Counsel Health
1wI’ve always respected “tumor board” for complex oncology cases - a multidisciplinary group (including medical oncology, surgical onc, rad-onc and others) gets together to discuss the nuanced tradeoffs for the most hairy, difficult cases. But time is limited, and the tumor board list can only be so long. What if every patient with multidisciplinary needs could get the benefit of all perspectives?